experimental realization of on-chip topological ... · 1 experimental realization of on-chip...
TRANSCRIPT
1
Experimental realization of on-chip topological nanoelectromechanical
metamaterials
Authors: Jinwoong Cha1,2, Kun Woo Kim3, Chiara Daraio2*
Affiliations:
1Department of Mechanical and Process Engineering, ETH Zurich, Switzerland
2Engineering and Applied Science, California Institute of Technology, Pasadena, CA, USA
3Korea Institute for Advanced Study, Seoul, Republic of Korea
*Correspondence to: Prof. Chiara Daraio ([email protected])
Topological mechanical metamaterials translate condensed matter phenomena,
like non-reciprocity and robustness to defects, into classical platforms1,2. At small
scales, topological nanoelectromechanical metamaterials (NEMM) can enable the
realization of on-chip acoustic components, like unidirectional waveguides and
compact delay-lines for mobile devices. Here, we report the experimental realization
of NEMM phononic topological insulators, consisting of two-dimensional arrays of
free-standing silicon nitride (SiN) nanomembranes that operate at high frequencies
(10-20 MHz). We experimentally demonstrate the presence of edge states, by
characterizing their localization and Dirac cone-like frequency dispersion. Our
topological waveguides also exhibit robustness to waveguide distortions and
2
pseudospin-dependent transport. The suggested devices open wide opportunities to
develop functional acoustic systems for high-frequency signal processing applications.
Wave-guiding through a stable physical channel is strongly desired for reliable
information transport in on-chip devices. However, energy transport in high-frequency
mechanical systems, for example based on microscale phononic devices, is particularly
sensitive to defects and sharp turns because of back-scattering and losses3. Two-dimensional
topological insulators, first described as quantum spin hall insulators (QSHE) in condensed
matter4-7, demonstrated robustness and spin-dependent energy transport along materials’
boundaries and interfaces. Translating these properties in the classical domain offers
opportunity for scaling the size of acoustic components to on-chip device levels.
Topological phenomena have been shown in various architectured materials like
photonic8-11, acoustic12,13, and mechanical1,2,14 - 18 metamaterials. Especially, photonic systems
have recently demonstrated the use of topological effects for lasing19-21 and quantum
interfaces22. However, acoustic and mechanical topological systems have so far been realized
only in large-scale systems, like arrays of pendula1, gyroscopic lattices2, and arrays of steel
rods12, laser-cut plates18 which require external driving systems. To fulfil their potential in
device applications, mechanical topological systems need to be scaled to on-chip level for high-
frequency transport.
Nanoelectromechanical systems23,24 (NEMS) can be employed to build on-chip
topological acoustic devices, thanks to their ability to transduce electrical signal into
mechanical motion, which is essential in applications. NEMS systems with few degrees of
freedom have already demonstrated quantum-analogous phenomena, like cooling and
amplification25, and Rabi-oscillation26,27. One-dimensional (1-D) nanoelectromechanical
lattices (NEML) are a different class of NEMS devices used to study lattice dynamics for
3
example, in waveguiding28,29, and energy focusing30. Recently, 1-D NEMLs made of SiN
nanomembranes have demonstrated active manipulation of phononic dispersion, leveraging
electrostatic softening effects and nonlinearity31.
Here, we fabricate and test high-frequency phononic topological insulators in
engineered, two-dimensional, nano-electromechanical metamaterials. We employ an extended
honeycomb lattice, which contains six lattice sites in a unit cell, satisfying C6 symmetry
(Supplementary information). This lattice exploits Brillouin zone-folding to demonstrate a
double-Dirac cone at point of the Brillouin zone (BZ). The zone-folding method has been
recently used in various topological photonic9,10, acoustic12,13 and elastic16 systems, by
introducing the concept of pseudospins that satisfy Kramers theorem7. BZ folding allows to
realize a pseudo-time reversal symmetry invariant system, where a time-reversal operator is
defined from the symmetry (C6) of the lattice. In our systems, the topological phase is controlled
by the coupling strength, t, between unit cells within the extended honeycomb supercell and
between adjacent cells, t’ (Fig. S1 and S2).
According to a classification method for topological phonons15, extended honeycomb
lattices are a class AII topological insulators, characterized by a Z2 topological invariant with
† 1T TU U . Here, TU is the time-reversal operator, which is anti-unitary (Supplementary
information). To calculate the topological invariant, we consider spin-Chern numbers from
BHZ model6 near the point, as pseudospins are not fully preserved in entire BZ. The spin-
Chern numbers C± = 0 for t > t’, C± = ± 1 for t < t’ confirms the pseudospin-dependent edge
states (Supplementary information). The Z2 topological invariants are = (C+ - C-)/2 (mod 2) =
0 for t > t’, and = 1 for t < t’, and support different topological phases. The difference in the
couplings, t and t’, leads to a band gap opening at point, with a dipole-to-quadrupole
vibrational band inversion for a non-trivial lattice (t > t’) (Supplementary information). If a
4
domain wall is formed between the two topologically distinct phases, gapless topological edge
states emerge.
We realize these topological properties in our NEMM by periodically arranging etch
holes, with 500 nm diameter, in an extended honeycomb lattice. The etch holes enable a
buffered oxide etchant (BOE) to partially remove the sacrificial thermal oxide layer and release
the SiN suspended membranes (Fig. 1a). We engineer the topological phases of the lattice, by
changing the distance between etch holes, w. This strategy allows us to control the lattice
couplings, t and t’, from the overlaps between the circular etching paths from the neighboring
etch holes (Extended Data Fig. 1). Recent reports28-31 on 1-D NEML have exploited the
isotropic nature of HF etching for device fabrication. Our NEMM consists of a periodic array
of free-standing SiN nanomembrane forming a flexural phononic crystal. The average thickness
of the nano-membranes is ~79 nm. The average vacuum gap distance between the SiN layer
and the highly doped silicon substrate is ~147 nm. These values are estimated considering the
partial etching rate of the SiN in the BOE etchant (~0.3 nm/min).
We perform finite element simulations using COMSOL©, to numerically compute
frequency dispersion curves for a unit cell with a lattice parameter, a = 18 m (Extended
Data Fig. 2). We vary the distance between two neighboring holes, w, from 5.5 m to 6.5
m (Fig. 1b to 1e). For a unit cell with w = 6.0 m = a / 3, a double Dirac-cone is present
around 14.55 MHz at the point of the BZ. (Fig. 1c). The frequency dispersion curves
typically start from around 12 MHz, because of the presence of clamped boundaries. The
frequency dispersion curves for w = 5.5 m and 6.5 m show the emergence of ~1.8 MHz-
wide band gaps at the point, ranging from 14 MHz to 15.8 MHz. The lattice with w =
5.5 m exhibits two additional band gaps below and above the center band gap around 15
MHz (Fig. 1b), while the lattice with w = 6.5 m (Fig. 1d) does not. The four vibrational
5
modes, px, py, dxy, and dx2-y
2, at the point are degenerate at the Dirac point for the lattice
with w = 6 m (Fig. 1c and 1e). The four degenerate modes are split into two separate
degenerate modes, for w < 6 m and w > 6 m (Fig. 1e), opening a band gap. The band
inversion between the dipole vibrational modes (px, py) and the quadrupole ones (dxy, dx2-
y2) appears at the point for w > 6 m, which supports the topological non-triviality of the
lattice.
To study the topological properties of our NEMMs, we first fabricate a straight
topological edge waveguide (Fig. 2a and 2b), formed at the interface of the topologically
trivial (w = 5.5 m, Fig. 2c) and non-trivial (w = 6.5 m, Fig. 2d) lattices. Topological
edge states do not exist at free-boundaries of our systems, owing to the lack of C6 symmetry.
The number of unit cells of each phase is approximately 200, so that the edge waveguide
has 20 supercells with 18 m one-dimensional lattice spacing. To characterize the edge
states, we excite the flexural motion of the membranes by applying a dynamic electrostatic
force, F (VDC + VAC)2, to the excitation electrode. Here, VDC and VAC are DC and AC
voltages, which are simulteneously applied between the excitation electrode and the
grounded substrate (Fig. 2a). We perform measurements using a home-built Michelson
interferometer with a balanced homodyne detection scheme (Methods). To obtain the
dispersion curves of the edge states, we measure the frequency responses of 20 sites along
the edge waveguide, by spatially scanning the measurement points (yellow strip, Fig. 2a)
with a 18 m step size (Extended Data Fig. 3). The Dirac-like edge state frequency
disperison curves, isolated from the bulk dispersion, are present in the frequency range
between 14.1 MHz and 15.8 MHz, showing a good agreement with the numerical
dispersion curves (Fig. 2f). We also observe a defect mode at the crossing point of the edge
state dispersion curves (Fig. 2e). This stems from a point defect mode from the boundary
6
near the excitation region. The broken C6 symmetry at the interfaces introduces a small
band gap in the middle of the edge state dispersions (Fig. 2b). Despite the presence of the
band gap, the defect mode is allowed to transmit non-negligible energy to the end of the
waveguide owing to the long decay length of the evanescent mode.
We also characterize the localization of the edge states, by scanning the
measurement point across the waveguide (yellow dashed-line AB in Fig. 2a) also with an
18 m step size. The edge states are strongly localized (Fig. 2h) within ± 36 m distance
from the interface (Fig. 2g). Beyond this range, the frequency responses (Fig. 2g) start to
show clear band gaps with similar positions and magnitude to the numerical dispersions,
shown in Fig1b and 1d. The trivial lattice side presents three band gaps (Fig. 2g, left) and
the non-trivial lattice side shows only one topological band gap (Fig. 2g, right), as predicted
in the numerical frequency dispersion (Fig. 1b to 1d). The frequency responses show
evidence of different topological phases in the two lattices, w = 5.5 and 6.5 m, confirming
that the waveguiding effect is topological.
One remarkable feature of topological edge modes is their robustness to waveguide
imperfections, like sharp corners. To study this, we additionaly fabricate a zig-zag
waveguide with two sharp corners, with 60o angles (Fig. 3a). We perform steady-state and
transient response measurements, by scanning the laser spot along the waveguide with 18
m scanning steps. The frequency dispersion from the steady state measurements confirms
the presence of the topological edge states (Extended Data Fig. 4). To validate the back-
scattering immunity, we perform transient response measurements with propagating pulses,
because the steady-state responses include signals from boundary-scattered waves. For
transient measurements, we apply a DC (VDC) and a chirped (VP) voltages to the excitation
electrode (Fig. 3a). A pulse with 14.25 MHz center-frequency and 0.2 MHz bandwidth,
propages at 77 m/s group velocity. The space-time evolutions of the pulse show a stable
7
energy transport with negligible backscattering from the two sharp corners (Fig. 3b). These
results demonstrate robustness of the waveguides.
Another crucial aspect of topological insulators is the unidirectional propagation for
distinct pseudospin modes. To characterize this, we fabricate another NEMM with a spin-
splitter configuration consisting of four domain walls, which has been employed in several
previous studies11,12 (Fig. 4a and 4b). Such geometry allows using a simpler pseudospin
selective excitation. In this configuration, the propagating direction of a pseudospin state
depends on the spatial configuration of the two topological phases, w = 5.5 and 6.5 m
(Fig. 4a). The pseudospin states are filtered to have a single dominant state in the input port
(yellow arrow in Fig. 4a). After the signal passes the input channel, the filtered spin state
mainly propagates to output port 1 and 3 (yellow arrows in Fig. 4a). We do not observe
propagation to port 2 since that channel does not preserve the incident pseudospin state in
the propagation direction (cyan arrow, Fig. 4a). To systematically investigate such
propagation behaviors, we send voltage pulses to the excitation electrode and measure
transient responses of the propagating pulses (Methods). We scan 13 sites (7 sites from the
input channel and 6 sites from each output channel) near the channels’ crossing point (Fig.
4b). Note that steady-state frequency response at the end of the three output ports (Fig. 4a)
exhibit almost identical edge state responses due to boundary scatterings (Extended Data
Fig. 5). The pulse we investigate has a 15.1 MHz center-frequency and 0.5 MHz bandwidth,
which is enough to cover the broad frequency ranges of edge states. As expected, the
measured signals show substantial energy transport to output 1 (Fig. 4c) and 3 (Fig. 4e),
but not to output 2 (Fig. 4d), after the crossing point (white arrows in Fig. 4c to 4e). The
presented results confirm that the propagation direction depends on the types of
pseudospins. The use of such spin selective excitation and detection methods will enable
compact, mechanical uni-directional devices.
8
The results we present in this work show that nanoelectromechanical metamaterials
can be used as platforms for on-chip topological acoustic devices. In the future, these
systems can be employed for stable ultrasound and radio frequency signal processing. With
advanced nanofabrication techniques, more sophisticated structures can be realized to
design other types of topological devices, for example, based on perturbative metamaterials
design methods17,18. Moreover, frequency tunability in nanoelectromechanical resonators
via electrostatic forces32,33 can be used for electrically tunable devices31 and actively
reconfigurable topological channel11.
Acknowledgements
We acknowledge partial support for this project from NSF EFRI Award No. 1741565, and the
Kavli Nanoscience Institute at Caltech.
Author contributions
J.C. and C.D. conceived the idea of the research. J.C. designed and fabricated the samples.
J.C. built the experimental setups and performed the measurements. J.C. performed the
numerical simulations. J.C. and K.K performed theoretical studies. J.C., K.K. and C.D. wrote
the manuscript.
Competing Financial Interests
Nothing to report.
9
Methods
Sample fabrication
The fabrication process begins with a pattern transfer by electron beam lithography and
development of a PMMA resist in a MIBK:IPA=1:3 solution. The excitation electrodes,
made of a Au (45 nm)/ Cr (5 nm) layer, are deposited on a 100nm-LPCVD silicon nitride
(SiNx)/140 nm-thermal SiO2/525 m highly-doped Si wafer, followed by a lift-off process
in acetone. A second electron beam lithography step, with ZEP 520 e-beam resist, is then
performed to create the pattern of etch holes (with 500 nm diameters) arranged in the
extended honeycomb lattices (Fig. 1a and Extended Data Fig. 1). We use an ICP-reactive
ion etch, to drill the holes on the SiNx layer. After we finish the etching of the holes, we
immerse the samples in a Buffered Oxide Etchant (BOE) solution for ~ 45 - 46 minutes, to
partially etch the thermal SiO2 underneath the SiNx device layer. The etching duration
determines the diameter of the etching circles, r (Extended Data Fig.1). Detailed
fabrication methods can be found in Ref. 31.
Experiments
The flexural motions of the membranes are measured using a home-built optical
interferometer (HeNe-laser, 633 nm wavelength) with a balanced homodyne method. The
measurements are performed at room temperature and a vacuum pressure P < 10-6 mbar. The
optical path length difference between the reference and the sample arms are stabilized by
actuating a reference mirror. This mirror is mounted on a piezoelectric actuator which is
controlled by a PID controller. The motion of the membranes is electrostatically excited by
simultaneously applying DC and time-varying voltages through a bias tee (Mini-circuits,
ZFBT-6GW+). The intensity of the interfered light from the reference mirror and the sample is
10
measured using a balanced photodetector, which is connected to a high-frequency lock-in
amplifier (Zurich instrument, UHFLI). The measurement position, monitored via a CMOS
camera, can be controlled by moving a vacuum chamber mounted on a motorized XY linear
stage.
For the dispersion curve measurements in Figure 2, we measure (at steady-state)
frequency responses ranging from 10 to 20 MHz of 20 scanned sites along the edge waveguide.
The scanning step is the one-dimensional lattice spacing, a = 18 m. The lock-in amplifier
(Zurich instrument, UHFLI) allows measuring the amplitude responses and the phase
differences between the measured signal and the excitation source. To plot the frequency
dispersion, we perform Fast-Fourier transformation of the amplitude sin(phase) data. The
amplitude only data and the phase-considered data are shown in Extended Data Figure 3a
and 3b.
For transient measurements in Figure 3 and 4, we send a chirped signal (AWG module
in UHFLI) and measure the signal with an oscilloscope (Tektronix, DPO3034). As the signal is
invisible for a low excitation amplitude, we first filter the RF-output signals from the
photodetector with a passive band-pass filter (6 – 22 MHz bandwidth) and average 512 data n
time-domain. For robustness measurements (Fig. 3), we send a pulse containing frequency
content ranging from 14 to 15 MHz, by applying VDC = 15 V and VP = 75 mV to the excitation
electrode. We then perform post-signal processing to extract signals of interest, by applying
a Burtterworth filter with 14.25 MHz center-frequency and 0.2 MHz bandwidth. For
pseudospin-dependent transport measurements, we use a pulse (14 – 16 MHz) and applied
VDC = 15 V and VP = 22.5 mV. We then apply a Burtterworth filter with 15.2 MHz center-
frequency and 0.5 MHz bandwidth.
11
Numerical Simulations
We perform finite-element simulations to calculate the phononic frequency dispersion
curves using COMSOL multiphysics. We employ the pre-stressed eigenfrequency analysis
module in membrane mechanics. We also consider geometric nonlinearity, to reflect the effects
of residual stresses. The physical properties of SiNx used in the simulations are 3000 kg/m3
density, 290 GPa Young’s modulus, 0.27 Poisson ratio, and 50 MPa isotropic in-place
residual stress. The lattice parameter, a, is chosen to be 18 m. We calculate frequency
dispersion curves for various unit cell geometries with different w ranging from 5.5 m to
6.5 m. The center hexagon and the six corners of each unit cell are fixed, due to the
presence of unetched SiO2 (light grey regions in the scanning microscope images in Fig.
2b-d). The radii of the etched circles are set to r = 4.9 m (Fig. 1 and Extended Data Fig.
2). We apply Bloch periodic conditions to the six sides of a unit cell,
( ) ( )exp( )u r R u r iq R , via Floquet periodicity in COMSOL. Here, r is a position
within a unit cell, R is a lattice translation vector, and q is a wave vector. We calculate
the dispersion curves along the boundary of the irreducible Brilluoin zone MA A KA (Fig.
S1b)
We also numerically calculate the frequency dispersion curves of the edge states to
validate the topological behaviors. As we are interested in one-dimensional dispersion
along the interface, we build a strip-like super cell with 18 m periodicity. Each topological
phase (w = 6.0 ± 0.5 m) spans about ± 160 m from the interface in the direction
perpendicular to the interface. We calculate the frequency dispersion by applying one-
dimensional Floquet periodic condition.
12
Data availability. The data that support the findings of this study are available from the
corresponding author upon reasonable request.
References
1. Süsstrunk, R., & Huber, S. D. Observation of phononic helical edge states in a
mechanical topological insulator. Science 349, 47-50 (2015)
2. Nash, L. M., et al. Topological mechanics of gyroscopic metamaterials. Proc. Natl.
Acad. Sci. 112, 14495-14500 (2015)
3. Baboly, M. G., et al. Demonstration of acoustic waveguiding and tight bending in
phononic crystals. Appl. Phys. Lett. 109, 183504 (2016)
4. Kane, C. L., & Mele, E. J. Quantum spin hall effect in graphene. Phys. Rev. Lett. 95,
226801 (2005)
5. Kane, C. L., & Mele, E. J. Z2 Topological order and the quantum spin hall effect. Phys.
Rev. Lett. 95, 146802 (2005)
6. Bernevig, B. A., Hughes, T. L., & Zhang, S. C. Quantum spin hall effect and topological
phase transition in HgTe quantum wells. Science. 314, 1757-1761 (2006)
7. Hasan, M. Z., & Kane. C. L. Colloquium: Topological insulators, Rev. Mod. Phys. 82,
3045-3067 (2010)
8. Khanikaev. A. B., et al. Photonic topological insulators. Nat. Mat. 12, 233-239 (2012).
9. Wu, L., Hu, X., Scheme for achieving a topological photonic crystal by using dielectric
material. Phys. Rev. Lett. 114, 223901 (2015)
10. Yves, S., et al. Crystalline metamaterials for topological properties at subwavelength
scales. Nat. Commun. 8, 16023 (2017)
11. Cheng, X., et al. Robust reconfigurable electromagnetic pathways within a photonic
topological insulator. Nat. Mat. 15, 542-548 (2016)
12. He, C., et al. Acoustic topological insulator and robust one-way sound transport. Nat.
Phys. 12, 1124-1129 (2016).
13. Yves, S., Fleury, R., Lemoult, F., Fink, M., & Lerosey, G. Topological acoustic
polaritons: robust and sound manipulation at the subwavelength scale. New J. Phys. 19,
075003 (2017).
14. Mousavi, S. H., Khanikaev, A. B., & Wang, Z. Topologically protected elastic waves
in phononic metamaterials. Nat. Comm. 6, 8682 (2015)
15. Süsstrunk, R., & Huber, S. D. Classification of topological phonons in linear mechanical
metamaterials. Proc. Natl. Acad. Sci. 112, E4767-E4775 (2016)
16. Chaunsali, R., Chen, C., & Yang, J., Subwavelength and directional control of flexural
waves in zone-folding induced topological plates. Phys. Rev. B 97, 054307 (2018)
17. Matlack. K. H., Serra-Garcia, M., Palermo, A., Huber, S. D., & Daraio, C. Designing
perturbative metamaterials from discrete models. Nat. Mat. 17, 323-328 (2018)
18. Serra-Garcia, M., et al. Observation of a phononic quadrupole topological insulator.
Nature 555, 342-345 (2018)
19. Bahari, B., et al. Nonreciprocal lasing in topological cavities of arbitrary geometries.
Science 358, 636-640 (2017)
20. Harari, G., et al. Topological insulator laser: Theory. Science 359, eaar4003 (2018)
13
21. Bandres, M. A., et al. Topological insulator laser: Experiments. Science 359, eaar4005
(2018)
22. Barik, S., et al. A topological quantum optics interface. Science 359, 666-668 (2018)
23. Huang, X. M. H., Zorman, C. A., Mehregany, M., Roukes, M. L.
Nanoelectromechanical systems: Nanodevice motion at microwave frequencies, Nature
421, 496 (2003)
24. Craighead, H. G., Nanoelectromechanical systems. Science 290, 1532-1535 (2000)
25. Alba, R. D., et al. Tunable phonon-cavity coupling in graphene membranes. Nat.
Nanotechnol. 11, 741-746 (2016)
26. Faust, T., Rieger, J., Seitner, M. J., Kotthaus, J. P., & Weig, E. M. Coherent control of
a classical nanomechanical two-level system Nat. Phys. 9 485-488 (2013)
27. Okamoto, H., et al. Coherent phonon manipulation in coupled mechanical resonators.
Nat. Phys. 9 485-488 (2013)
28. Hatanaka, D., Mahboob, I., Onomitsud, K., & Yamaguchi, H. Phonon waveguides for
electromechanical circuits. Nat. Nanotechnol. 9, 520-524 (2014)
29. Hatanaka, D., Dodel, A., Mahboob, I., Onomitsud, K., & Yamaguchi, H. Phonon
propagation dynamics in band-edngineered one-dimensional phononic crystal
waveguides. New J. Phys. 17, 113032 (2015)
30. Kurosu, M., Hatanaka, D., Onomitsu, K., & Yamaguchi, H., On-chip temporal focusing
of elastic waves in a phononic crystal waveguide. Nat. Commnun. 9, 1331 (2018)
31. Cha, J., & Daraio, C. Nonlinear nano-electromechanical lattices for high-frequency,
tunable stress propagation. Preprint at https://arxiv.org/abs/1803.06427 (2018)
32. Chen, C., et al. Graphene mechanical oscillators with tunable frequency. Nat. Nanotech.
8, 923-927 (2013)
33. Bagci, T., et al. Optical detection of radio waves through a nanomechanical transducer.
Nature 507, 81-85
14
Figure 1. Unit cell geometry and topological phase transitions. (a) Schematic of a two-
dimensional nanoelectromechanical metamaterial. The grey area represents the silicon nitride
nanomembrane suspended over a highly doped n-type silicon substrate. The black dots, forming
a honeycomb lattice, represent etch holes. The light blue hexagons represent the unetched
thermal oxide, acting as fixed boundaries. The unit cell geometry (black solid hexagon) is
shown in the right inset, with relevant parameters. An example flexural mode is shown in the
left inset. The topological phases are controlled by changing w. (b), (c), (d) Frequency
dispersion curves along a boundary of the irreducible Brillouin-zone MKM, when (b) w = 5.5
m, (c), 6.0 m, and (d) 6.5 m. The red and green shaded regions correspond to topological
and non-topological band gaps, respectively. (e) Eigenfrequencies above and below the
topological band gap at the point, as a function of w. Blue (red) dots denote the
eigenfrequencies for flexural modes px and py (dxy and dx2
-y2). The flexural mode shapes are
presented for w = 5.5 m (left), 6.0 m (middle), and 6.5 m, respectively (right).
15
Figure 2. Characterization of topological edge states. (a) Scanning electron microscope
image of a straight topological edge waveguide. The two different topological phases are shaded
by blue (non-trivial) and red (trivial) false colors. Flexural membrane motions are excited by
simultaneously applying DC and AC voltages (VDC = 2 V, VAC = 20 mV) to the excitation
electrode via a bias tee. Scale bar, 100 m. (b), (c), (d) Scanning electron microscope images
of (b) edge region, yellow shaded strip in (a), (c) trivial lattice with w = 5.5 m, the red shaded
area in (a), and (c) non-trivial lattice with w = 6.5 m, the blue shaded area in (a). Scale bars
are 10 m. The red and blue dots in (b) denote the lattice points for w = 5.5 m, and w = 6.5
m, respectively. The blue and red hexagons in (c) and (d) represent the unit cells for w = 5.5
m, and w = 6.5 m. (e) Experimental and (f) numerical frequency dispersion curves along the
edge waveguide (from C to D in a). Yellow and light blue dots in the edge state dispersion in
(f) represent propagating waves for up- and down- pseudospins, respectively. Time-evolution
of the mode shapes at points 1,2,3, and 4 are provided in supplementary video files. (g)
Frequency responses for 19 different sites along line A-B shown in (a) (middle panel). The left
and right panels represent frequency responses at site A and B, respectively. The red and green
shaded regions represent the band gaps. (h) Flexural modes for point A and B in the numerical
dispersion shown in f. The width of the strip is 18 m, identical to the lattice parameter a.
16
Figure 3. Waveguide robustness against imperfections. (a) Scanning electron microscope
picture of a zig-zag topological edge waveguide. The red and blue shaded regions represent
topologically trivial and non-trivial lattices, respectively. The flexural membrane motions are
excited by simultaneously applying DC and a chirped signal with frequency content ranging
from 14 MHz to 15 MHz. The applied voltages are VDC = 15 V, VP = 75 mV, here VP is the
amplitude of the chirped signal. The orange dots represent measurement points. Scale bar is
100 m. (b) Transient responses along the edge waveguide in a space-time domain. A pulse
with 14.23 MHz center-frequency and 0.2 MHz bandwidth is considered. The position denotes
the measurement points in the scanning direction shown in (a). c1 and c2 mark the position of
the sharp corners. The solid line is added to highlight the trajectory of the propagating pulses.
The two dotted lines mark the expected trajectory of the backscattered signal, if the propagating
pulse were reflected from corners c1 and c2.
17
Figure 4. Pseudospin-dependent wave propagation. (a) Scanning electron microscope image
of a pseudospin-filter configuration. Scale bar is 100 m. The two different topological phases
are shaded by red (trivial, w = 5.5 m) and blue (non-trivial, w = 6.5 m) false colors. Flexural
motions are excited from the electrode (VDC = 15 V, VP = 22.5 mV). The yellow and cyan arrows
represent propagating directions for different pseudospin states. (b) Zoomed-in region marked
by the dotted square in (a). The red and blue dots denote the lattice points for trivial and non-
trivial phases. Scale bar is 30 m. (c), (d), and (e) Envelopes of propagating pulses (15.2 MHz
center-frequency, 0.5 MHz bandwidth) in space-time domain. The position represents the
measurement points along the edge waveguides. (c) Input to Output port 1, (d) Input to Output
port 2 (e) Input to Output port 3. The crossing point is indicated by white arrows in (c), (d), and
(e).